Research & Frameworks

Building the Science of
AI Governance

Research-backed frameworks and methodologies for enterprise AI compliance, governance, and risk management. Built from real-world implementation experience.

Core Frameworks

Research Contributions

Frameworks and methodologies developed through years of enterprise AI implementation. Each addresses critical gaps in traditional approaches to AI governance.

Precision Drift Detector

Active Research
2024-2025
Advanced methodology for detecting subtle degradation patterns in production AI systems. Goes beyond basic statistical drift to identify concept drift, performance degradation, and silent failures before they impact users.

Key Contributions

Multi-dimensional drift analysis framework
Early warning signal detection
Context-aware threshold adaptation
Production incident correlation

Cognitive Systems Management (CSM)

Published Framework · Current Version: CSM 2.0
2025-2026
A four-domain AI governance methodology comprising CSM-Enterprise, CSM-Project, CSM-Code and CSM-UX. First published August 29, 2025 in "Cognitive System Management: A Framework for Enterprise AI Project Governance" on AI Governance on Medium. Current version: CSM 2.0 - a deterministic-by-design governance operating model that preserves the original four domains and adds six cross-functional execution functions plus versioned governance contracts.

Key Contributions

Four-domain governance model (Enterprise, Project, Code, UX)
Cross-domain handoff visibility
Risk-integrated governance methodology
CSM 2.0: governance contracts, evidence schemas, reassessment triggers
Deterministic reference evaluator for objective rules
Practitioner-developed framework, versioned for public scrutiny

Red Audit Kit

Active Framework
2024-2025
Systematic assessment framework for AI systems covering models, data pipelines, infrastructure, and governance. Provides structured methodology for identifying compliance gaps and risk exposure.

Key Contributions

Multi-layer audit methodology
Risk scoring and prioritization
Regulatory mapping automation
Remediation roadmap generation

LegacyShift Methodology

Active Framework
2024-2025
Structured approach to modernizing legacy AI systems. Addresses technical debt, compliance gaps, and operational inefficiencies while minimizing risk and maintaining business continuity.

Key Contributions

Zero-downtime migration patterns
Incremental modernization strategy
Risk-managed transitions
Compliance preservation frameworks
Research Focus

Active Research Areas

AI Regulatory Compliance

  • EU AI Act implementation strategies
  • Cross-jurisdiction compliance frameworks
  • Automated compliance monitoring
  • Policy-to-implementation mapping

Enterprise AI Governance

  • Multi-model governance at scale
  • Organizational governance structures
  • Stakeholder alignment frameworks
  • Governance automation

AI Risk Management

  • Silent failure detection
  • Cascading risk analysis
  • Risk quantification methodologies
  • Real-time risk monitoring

System Modernization

  • Legacy AI migration patterns
  • Technical debt assessment
  • Modernization without disruption
  • Compliance-preserving refactoring
Publications

Published Work

The Instruction Stack Audit Framework (ISAF): A Technical Methodology for Tracing AI Accountability Across Nine Abstraction Layers

Addresses AI accountability failures by providing a nine-layer technical specification for documenting instruction propagation from hardware to outputs. Includes 127-checkpoint audit protocol, cryptographic verification, and risk scoring based on abstraction distance. Demonstrates application to EU AI Act, NIST AI RMF, and ISO/IEC 42001 compliance requirements.

AI GovernanceEU AI ActNIST AI RMF
Technical Report·Zenodo·2025
view paper →

Deterministic Bias Detection for NYC Local Law 144: Why Reproducibility Matters More Than Accuracy

Presents a reproducibility-first architecture for detecting linguistic bias in job descriptions under NYC Local Law 144. Argues that regulatory compliance requires deterministic systems over probabilistic ML models. Details rule-based pattern matching, version-controlled lexicons, reproducible scoring, and cryptographic evidence generation for legally defensible documentation.

Algorithmic AccountabilityBias DetectionDeterministic AI
Technical Report·SSRN·2026
view paper →

From Industrial Electrification to Artificial Intelligence: Institutional Lessons from Construction Governance for AI Risk Regulation

Analyzes the institutional evolution of construction governance and applies its structural lessons to AI risk regulation. Draws from historical developments in mechanization, electrification, occupational safety regulation, professional licensing, and insurance enforcement to propose a phased governance maturation model for AI systems.

AI Risk RegulationGovernance MaturationGeneral Purpose Technology
Working Paper·Zenodo·2025
view paper →

From AI Pilots to Regulatory Readiness

Practical framework for transitioning from AI experimentation to production-grade, compliant systems.

Framework Paper·AI Governance Playbook·2025
view paper →

Why Enterprise AI Integration Strategies Fail

Systematic analysis of common architectural and organizational failures in enterprise AI adoption.

Analysis·Design Bootcamp·2025
view paper →

Cognitive System Management: A Framework for Enterprise AI Project Governance

Original publication of the Cognitive Systems Management (CSM) framework: a four-domain governance methodology comprising Enterprise, Project, Code and UX. Current version: CSM 2.0 (spec v2.0.0, 2026-08-10).

Methodology·AI Governance on Medium·2025
view paper →
Collaboration

Research Partnerships

I collaborate with organizations on applied AI governance research. If you're working on problems in compliance automation, risk quantification, or governance at scale, let's talk.

Apply These Frameworks to Your Organization

These aren't just academic exercises. They're battle-tested frameworks built from real-world enterprise implementation. Available through HAIEC platform and advisory engagements.

Discuss AI →